I am attending InterACT San Francisco 2007 this week and blogging live (or nearly live) from the sessions. This session is "Rules-Based Product and Pricing Engine (Everbank)" presented by EverBank and their implementation partner ValueMomentum. The project was consolidation of pricing for all the front-end systems at a medium size bank.
Everbank have lots of different front-end systems, all supporting different channels but the product and pricing rules were all separate. Rolling out new products or making changes was really hard. Sometimes there was a paper based manual that differed from the systems actual code. Lots of product codes to work around things and so on. Wanted to decouple the business functionality for pricing and product selection and use a service-oriented approach to plug the resulting decision service into all the front-end system. One place, one set of edits, all systems impacted. Also wanted to engage the business in this process. It had to be possible to add new products by reusing pieces of existing products and do so quickly from a business perspective. This allowed different lines to share business know-how through reusable components.
The choice was between buying an application (and doing some customization) and building a custom product and pricing engine as a core competency, a core piece of infrastructure. They chose to use Blaze Advisor and build a core engine for product and pricing. Did some "best case" cash flows to compare value and costs of different approaches. The numbers for the business rules approach had an IRR of 40% and an ROI of 75%. They estimated a 50% reduction in staffing to do maintenance on pricing and estimated significant opportunity cost savings from faster time to work. Even their "worse case" scenario was pretty good.
They had failed to bring in a suitable engine several times thanks, in part, to a lack of collaboration. They addressed this by taking the consultants, architects, business users and sequestered them away from their work so that the whole team had a common goal. Focused on empowering this group to find the answers their needed. The project set up a very thorough quality approach so that testing took place as they went - had to show the same result as the existing systems. The project wanted to be sure that the new system was faster as well as better. Able to price 100 products for a loan in 10 seconds! They also incented the team with a separate (3x) bonus for the completion of the project!
They used a template driven approach, using prior experience that ValueMomentum had with rules. This allowed the IT folks to develop rule templates that could then power business user-friendly rule maintenance. The business needed the confidence that they could make changes in respond to market changes so the infrastructure supported nightly updates. The business users drove much of the rule maintenance inside this daily cycle, though using fairly flexible templates that give business users a lot of options. For instance, one group completely changed the rules for a line of business without needing technology support. Some specific challenges and approaches:
- Automated testing was a key component of making sure the system was as fast and reliable as necessary.
When building a central component that will be reused, these are critical issues. For instance there were 10,000 test cases for their 3,000 rules at go-live.
- All the pricing rules have to be managed for historical circumstances and effective dating.
Put in historical rules going back several months so that they could run in parallel as well as price the whole backlog when the system went live.
- Ran old system for the first few months
Gave them several months to prove the pricing was consistent between the systems. Then decided to stop maintaining old one as were convinced of the new systems stability.
- Needed real control of rules
Individual brokers need specific rules, detailed pricing and bonus rules for channels etc. Had version control and auditing enabled to ensure security and lifecycle control.
- 99.8% match but the new system was right 100%!
- Sub-second response time for a single product
Live for about 7-9 months now and have 4 of the 7 systems integrated. They have had no new channels yet nor have the seen a reduction in risk, largely thanks to the continuing use of the old approach in some systems for the moment. They have rolled out new regions and new products (business-driven, less than a week to do), price accuracy is high and it is much less effort to make changes. All the costs were as expected except that integration costs were higher. Here are a few slides:
A post on a platform for mortgage and another use of rules in the mortgage industry and some articles in Mortgage Technology - Rules-Based Systems Call the Plays and Are Business Rules Engines for Everybody? You can find the full set of posts from InterACT in this category.